Ticker Feed Builder

Downloads RSS feeds on a schedule and turns each one into numbered text lines and JPG images for TV tickers and playout graphics. Free desktop app (Electron) from OnAir Garage.

• Tool page: https://onairgarage.com/tools/ticker-feed-builder/
• Author: Graziano Melzi · OnAir Garage — [email protected]
• License: MIT (see LICENSE)
• Systems: Windows (installer), macOS (dmg, Apple silicon and Intel), Linux (AppImage, x64). Built for all three; Windows and macOS have been run by the author (declared 17 October 2026); Linux has only been run in the automated tests and the end-to-end run under a virtual screen (see Validation).
• Manual: attached to every release as three PDFs, English (Ticker-Feed-Builder-Manual-EN-<version>.pdf), Italian (-IT-) and Spanish (-ES-); the source is in docs/manual/ (python3 docs/manual/build.py <version> <folder> makes the PDFs; it needs Chrome or Chromium).

Screenshots use invented demo feeds, taken from the real app (version 26.10.9, running on Linux; dark and light themes).

What it does

It replaces the rss_ticker_downloader.py scripts that fed NeMedia Morpheus from Windows Task Scheduler. The app lives in the tray, runs 24/7 and, for every feed of every profile, writes:

File	Content
<Name>_Title.Txt	one headline per line
<Name>_Description.Txt	one description per line, same order
<Name>/00001.JPG, 00002.JPG…	one JPG per story, same order
<Name>_metadata.json	counters of the last run (optional)

• Profiles: one destination each (output folder, feeds, interval, placeholder image, format). Use several if different screens read different folders.
• Sync guarantee: story N is line N of both text files and image N. Line breaks and invisible characters inside a text are removed; an empty title/description becomes - (configurable); a missing or unreadable image becomes the placeholder image.
• Safe writes: each file is written under a temporary name and renamed; unchanged files are not touched; if a download fails the previous files stay as they were; a missing output folder is reported and never created (an unplugged drive must not become a local folder); numbered images left over from a longer previous run are removed.
• Configurable format (the defaults reproduce the original scripts): text encoding (UTF-8, UTF-8 with BOM, Windows-1252), line ending, file name templates ({folder}), image number start/digits/extension, JPEG quality, optional resize (fill or fit), title/description length limits.
• Checks: each profile has a check interval (default 5 min) and each feed can have its own. A check first asks the server whether the feed changed (ETag / Last-Modified; answer 304 = nothing is downloaded). If the server cannot say, the feed is downloaded and its content compared with the previous copy (same content = no image download, no file touched). Images are downloaded and files written only when something is new; a failed image is retried at the next check, and a deleted output file or an edited setting triggers a rebuild. "Run now" always rebuilds everything.
• Schedule of a profile, three ways: every N minutes after the previous check; every N minutes on the clock (counted from midnight: 15 = :00 :15 :30 :45, 60 = every full hour); or at fixed times on the days you choose (one time = once a day). At a fixed time every enabled feed is checked, whatever its own interval; a time missed while the app was closed is made up once at start-up (if the schedules start at launch); time windows always apply. Stagger the profiles (Settings, off by default) spreads profiles that share an interval evenly inside it instead of starting together (two profiles every 5 min: 2 min 30 s apart).
• Dashboard (its own button above the profile list): date and time, schedule state, feeds OK / with problems, a time-ordered list of the next checks (what will be checked and when, with countdown), per profile and per feed last check and last file update with date and time, recent warnings and errors, and for every feed a strip with the last 24 hours, one bar per hour (new stories / checked, nothing new / failed / no check), kept across restarts. Profiles show in the order of the left column and say in words if they are OK, partial or have a problem.

The default layout is the one the original scripts produced for NeMedia Morpheus. It has not been checked against NeMedia documentation (none was read); other ticker systems may need other names, encodings or sizes, which is why everything is configurable.

Feeds and publishers' terms

A feed that is free to download is not necessarily free to broadcast. Publishers usually offer RSS for personal reading; showing their headlines publicly on TV, screens or websites normally needs an agreement with the publisher. That responsibility is the user's; this is not legal advice. The app ships without any feed; the examples below come from the pages read on 8 October 2026 (paraphrased; the pages carry no version or date, so read the current text before relying on it).

Publisher	What its RSS page says	Page
ANSA	Non-commercial use by individuals and non-profit organisations, only to view the feeds in RSS reader programs; using them to publish the latest headlines on websites or blogs is not allowed; any other use must be requested and expressly authorised.	https://www.ansa.it/sito/static/ansa_rss.html
Adnkronos	Non-commercial use by individuals and non-profit organisations; the feeds may not be made public (websites, blogs, outdoor communication systems, any other means that spreads the news publicly) without a prior agreement with Adnkronos.	https://www.adnkronos.com/rss

Both pages list their feeds by section (politics, foreign news, sport…); use them as examples of what to ask for in an agreement.

Formulas and sources

No measurements or standards-based formulas. The external rules used:

Value / rule	Source	Status
RSS 2.0 item fields (title, description, enclosure), Atom entry, Media RSS media:content / media:thumbnail	feed formats as found in the feeds themselves (Adnkronos, ANSA); no specification was read for this release	recommended (practice, not verified against the specifications)
Characters not allowed in a file name, reserved names (CON, PRN, NUL, COM1…)	Microsoft Learn, Naming Files, Paths, and Namespaces — same rules as in File Renamer	official (read in the File Renamer project, not re-read for this release)
Default output layout and file names	original scripts v3.2 supplied by the author	reference implementation
Telegram sendMessage	Telegram Bot API	not re-read for this release
SMTP submission (STARTTLS / TLS), Bcc	nodemailer library defaults; no RFC read for this release	recommended (practice)
Terms of use of the ANSA and Adnkronos RSS feeds	the two RSS pages above	official (read 2026-10-08, summarised)

Assumptions and limits

• Images are decoded without native modules: JPEG, PNG, GIF, BMP and TIFF with a JavaScript library (jimp), WebP and AVIF with WebAssembly decoders (@jsquash, Apache-2.0). A very large AVIF can take a few seconds; an image that cannot be read gets the placeholder.
• Images on private network addresses (localhost, 10.x, 172.16–31.x, 192.168.x…) are ignored for feeds on the public internet; DNS names that resolve to private addresses are not detected.
• "Skip certificate check" keeps the connection encrypted but does not verify the server; it is off by default and meant only for feeds you trust (some publishers serve their feeds with a broken certificate).
• The computer must stay on and awake. The app asks the system not to sleep (option), but cannot stop a forced sleep or shutdown.
• Whether a server supports conditional requests depends on the server; without it every check downloads the feed XML (small) but still skips images and writes when the content is identical. Stories are taken in the order of the feed; there is no de-duplication across feeds.

Validation and references

Results as of October 2026.

Reference	What it validates	Result	How to re-run
Automated tests (82)	text cleaning, text changes (replacements, capital letters, text before / after, own limits), the ad-hoc signature hook, RSS/Atom parsing, line/image sync, empty values, orphan removal, unchanged files, change detection (304, same content, failed-image retry, deleted output, per-feed interval), WebP/AVIF decoding, time windows, filters, merged feeds, duplicates across feeds, file check, remembered state, frozen-feed alert, daily summary, headless CLI, failure keeps old files, missing folder not created, private image addresses, scheduler, alerts	pass	npm test
Original Python script v3.2 on 8 live feeds (2026-10-08, macOS; a one-off development check with the feeds the script used, not part of the app)	same titles and descriptions as the original for the same feeds	14 of 16 text files byte-identical; the other 2 differ only because the feeds changed between the two runs (stories shifted by one); one title carried an invisible BOM character in the original, which the app removes	run both on the same feeds and compare (cmp)
Same run	number of images per feed	equal for all 8 feeds	same
End-to-end on the real Electron app (macOS; Linux for the 26.10.10 interface)	create profile, run, files on disk, status in the UI, language switch, Settings tabs, show / hide password, backup box, Help & About, feed options, no console errors	pass	node dev/e2e-electron.mjs [screenshotDir]

Not validated: WebP/AVIF with real-world files from publishers (only synthetic images in the tests); the headless mode on Windows and as a service; playback in NeMedia Morpheus or any other ticker system; Linux on a real desktop; runs lasting days; Telegram and email delivery against real services (unit-tested with fakes; the email envelope is checked with the real Nodemailer message builder, not with a mail server); Windows-1252 output beyond a unit test. Opening the ad-hoc signed app on an Apple Silicon Mac (only the build job checks the signature; the author tries the preview by hand); a click on the tip jar opening the system browser on Windows and macOS; the text changes on a live ticker. This is not a certified tool.

Sources to re-check

Document	Version read
RSS 2.0 specification, Atom (RFC 4287), Media RSS	not read for this release
Telegram Bot API	not read for this release
SMTP (RFC 5321 / 6409)	not read for this release
NeMedia Morpheus manual (expected file layout)	not read
ANSA RSS page (ansa.it/sito/static/ansa_rss.html)	read 2026-10-08, no version or date on the page
Adnkronos RSS page (adnkronos.com/rss)	read 2026-10-08, no version on the page

Run locally

npm install
npm start          # the app
npm test           # unit tests (Node 20+)
node dev/e2e-electron.mjs   # end-to-end on the real app
npm run dist:mac   # or dist:win / dist:linux

Headless mode

src/cli.js runs the same engine, schedules, filters, alerts (Telegram, email) and daily summary without a window or tray (no desktop notifications).

node src/cli.js --settings exported-settings.json --data ./tfb-data          # runs until stopped (Ctrl+C)
node src/cli.js --settings exported-settings.json --data ./tfb-data --once   # one run, then exit

The installed app can run it without Node.js: ELECTRON_RUN_AS_NODE=1 "<app executable>" "<resources>/app.asar/src/cli.js" --settings … --data … (the exact command for this computer is shown in Settings → Headless mode). --once exits with 0 (all feeds fine), 1 (some problem) or 2 (settings unusable). Use the file written by "Export settings"; it contains no secrets, so set TFB_TELEGRAM_TOKEN and TFB_SMTP_PASS in the environment. The file is read again every 30 seconds. On Windows start it with Task Scheduler at startup ("run whether the user is logged on or not"), on Linux with a systemd service, on macOS with launchd. Logs and the remembered state go to the --data folder.

Structure

• main.js, preload.cjs — Electron main process and bridge.
• src/ — the engine, independent of Electron: engine.js (download → filter → convert → write → check), runtime.js (engine + scheduler + alerts + summary + remembered state, shared by the app and the headless mode), cli.js, feed.js, filters.js, schedule-rules.js, digest.js, text.js, image.js, image-modern.js, output.js, format.js, http.js, scheduler.js, notify.js, messages.js, settings.js, logger.js, updater.js, secrets.js (the one list of every secret) and backup.js (the backup file and its import). updater.js, secrets.js and backup.js start with // garage-common: <name> <version>: they are copies of modules shared by the OnAir Garage apps.
• renderer/ — the interface (strict CSP, no inline scripts or styles, no CDN, system fonts).
• assets/ — the logo as SVG (logo.svg app icon, tray.svg single-colour menu-bar glyph), the generated icons and the built-in placeholder. node dev/make-icons.mjs regenerates every icon from the two SVGs (needs Playwright with a Chromium; see the top of the script).
• dev/ — tests, end-to-end script, icon and placeholder generators.

Versions are YY.M.N (e.g. 26.10.1), tags vYY.M.N.

Credits

fast-xml-parser (MIT), jimp (MIT), iconv-lite (MIT), nodemailer (MIT-0), @jsquash/webp and @jsquash/avif (Apache-2.0), Electron (MIT). No fonts are bundled.